基于振动与电机电流信息融合的转子系统载荷识别及故障诊断方法

The common features of large scale machinery equipment rotor systems are complicated structure, heavy load bearing, motor driving with large power, obverious vibrication facts and more factors affecting load and fault. Under the frequent excitation of load, new charactoristics of rotor systems vibrication and motor current will occur. Study on which shall release important significance to safety and reliable running of large scale machinery equipment. In the view of above facts, a new train of thought of rotor systems load identification and fault diagnosis based on information fusion technique of vibrication and motor current signals comes. Building system dynamics equation containing vibrication and motor current characteristic factor, revealing mapping relationship between load, fault and signals features of motor current and vibrication and carrying out qualitative and quantitative identification on different types of loads acting on rotor systems through information fusion can carry out diagnosis and earlier alarm to the possible malignant accidents. The study releases the abtaining channels of load excitation and signals of motor current and vibrication and the methods of coupling signal decompose and feature extraction. The obtained ways of effectively picking information of load and fault from aliasing signals by the test identification of load exciting and data mining of physical dynamic features of operation system would be capable of increasing accuracy of load identification and fault diagnosis. Verifying aforesaid theories through testing research will dedicate theoretical basis for dynamic design and fault diagnosis of rotor systems.

大型机械装备转子系统共同特点是结构复杂、承受载荷大、驱动电机功率大、振动问题突出、影响载荷与故障的因素多。在载荷频繁激励下，转子系统的振动和驱动电机电流将出现新的特征。研究这种变化特征，对大型机械装备安全可靠运行具有重要意义。 鉴于此，提出了基于振动与电机电流信息融合的转子系统载荷识别及故障诊断的新思路。建立含振动和电机电流特征因素的转子系统动力学方程，揭示载荷、故障、电机电流、振动特征间的映射关系；通过信息融合对转子系统承受的不同类型载荷进行定性与定量识别，对可能发生的恶性事故可进行诊断和早期预示。 研究给出了载荷激励与系统电机电流和振动信息的获取途径、耦合信号分解与特征提取方法；通过载荷识别试验和系统运行物理动态特征的数据挖掘，得到从混叠信号中有效提取载荷与故障信息方法，可提高载荷识别与故障诊断的准确性，通过试验验证上述理论的正确性，为转子系统动力学设计和故障诊断提供理论依据。

在机械工程领域，转子系统作为关键部件得到广泛应用。通常转子系统会在各种载荷激励下运行，因此对系统受到外载荷激励时的响应规律进行研究和探讨，进而进行载荷激励下的故障诊断研究就显得尤为重要。课题分别基于振动、电流等信息进行载荷辨识与故障诊断。. 推导出系统在典型载荷与故障激励下的的弯扭耦合、机电耦合方程；构建弯扭耦合与机电耦合仿真模型；仿真研究载荷、故障激励下的振动、电流与速度特性。.针对载荷识别，提出基于振动信息的EEMD能量化和BPNN分类筛选的载荷定性辨识方法，ELM方法定量回归拟合各载荷。运用SVD方法除去电流信号工频成分，通过WPA提取信号能量特征，利用LVQNN方法进行载荷分类并构建载荷定量辨识GRNN关系模型，反求出系统所受载荷。采用支持向量机进行基于速度的载荷辨识。提出了基于振动与电流信息的贝叶斯估计技术的载荷定性辨识方法，基于融合信息的定量辨识SVMR方法。辨识结果显示效果优良，满足实际辨识的需要。. 针对典型故障，采用形态学图像处理和Hu不变矩进行特征向量提取，应用BP神经网络进行轴心轨迹的自动识别。提出基于EEMD和平移不变多小波相邻系数的齿轮箱故障特征提取方法。利用变分模态分解、共振信号稀疏分解，将振动信号按频段分解，诊断不同故障类型。对电流信号进行经验模态分解，并采用IMF能量与IMF峭度相结合的方法提取特征向量，将其输入GA-SVM中进行故障类型识别。提取了电流和振动信号的小波包能量特征，运用主元分析法进行降维处理；设计贝叶斯网络，分别基于振动、电流信号、融合信息进行故障诊断。结果表明，融合信息故障诊断准确率高。.研制和搭建了满足该项目研究要求的转子系统变载荷试验台、提升机试验台、双跨交直流试验台、减速器多故障试验台和滑动轴承磨损试验台。. 通过本项目的研究，揭示了载荷、故障、振动、电流等之间的耦合关系；通过信息融合，完成了载荷识别与故障诊断。研究为机械装备转子系统动力学设计、运行监测、载荷辨识与故障诊断提供了理论依据和试验支持。